What is the difference between a martingale and a semi-martingale? Provide an example of each in the context of finance.?

In the context of probability theory and stochastic processes, martingales and semi-martingales are widely used concepts, particularly in quantitative finance for modeling price dynamics and risk management.

A **martingale** is a stochastic process (X_t)_t ≥ 0 which satisfies the following conditions:

1. Adaptedness: X_t is measurable with respect to the filtration ℱ_t for all t ≥ 0. This means that the value of the process at time t can be determined by the information available up to that time.

2. Integrability: 𝔼[|X_t|] < ∞ for all t ≥ 0. This condition requires that the expected value of the absolute value of the process is finite at each time point.

3. Martingale property: 𝔼[X_t + s|ℱ_t] = X_t for all s ≥ 0 and t ≥ 0. This means that the best predictor of the process at any future time, given the available information up to the current time, is its current value.

A **semi-martingale** is a stochastic process (Y_t)_t ≥ 0 that can be decomposed into the sum of a martingale process (M_t)_t ≥ 0 and a finite variation deterministic process (V_t)_t ≥ 0. In other words, there exist two processes M_t and V_t such that:

1. Y_t = M_t + V_t for all t ≥ 0.

2. M_t is a martingale.

3. V_t has finite variation on every interval [0, t].

In summary, every martingale is a semi-martingale, but not every semi-martingale is a martingale.

Now let’s provide an example in the context of finance.

**Martingale example:**

A typical example of a martingale in finance is the **simple symmetric random walk**, which can be viewed as a model for the price of a stock in a frictionless market with no predictable trends. Let S_t represent the stock price at time t, and let ΔS_t denote the (discrete-time) change in the stock price. We assume that ΔS_t takes values  ± 1 with equal probability, i.e.,

$$\mathbb{P}(\Delta S_t = 1) = \mathbb{P}(\Delta S_t = -1) = \frac{1}{2}.$$

Then, we can define the cumulative gain process $X_t = \sum_{i=1}^t \Delta S_i$, where X₀ = S₀. (X_t)_t ≥ 0 is a martingale because it satisfies the conditions mentioned above.

**Semi-martingale example:**

In finance, a well-known example of a semi-martingale is given by the **Black-Scholes model** for stock price dynamics. The model is described by the following stochastic differential equation (SDE):

dS_t = μS_tdt + σS_tdW_t,

where S_t represents the stock price at time t, μ is the drift (expected return) of the stock, σ is the volatility, and W_t is a standard Brownian motion (also known as a Wiener process).

We can decompose the stock price process S_t into a martingale and a deterministic process:

$$S_t = S_0\left(\exp\left(\sigma W_t\right) \exp\left(\left(\mu - \frac{\sigma^2}{2}\right)t\right)\right).$$

The first factor is a martingale:

M_t = S₀exp (σW_t).

The second factor is a deterministic process of finite variation:

$$V_t = S_0 \exp\left(\left(\mu - \frac{\sigma^2}{2}\right)t\right).$$

Thus, (S_t)_t ≥ 0 is a semi-martingale in the Black-Scholes model. Note that if the Black-Scholes model is modified to have zero drift (μ = 0), then the stock price process becomes a martingale, specifically a geometric Brownian motion.